一种基于压缩域的视频镜头检测方法  

Video Shot Detection Method Based on Compressed Domain

作　　者：于鹏[1] 王永滨[1] 柯雅明 刘文[1] 伏文龙[1] 

机构地区：[1]中国传媒大学计算机学院,北京100024

出　　处：《电视技术》2014年第13期7-10,共4页Video Engineering

基　　金：国家"863"计划项目(2011AA01A107);国家科技支撑计划项目(2012BAH51F02)

摘　　要：针对传统的基于空域的视频镜头检测算法需要解压缩,计算量大、效率低的缺点,提出了一种基于压缩域的视频镜头检测方法。该方法首先根据MPEG压缩标准,从视频流中提取I帧中8个低频DCT系数,并对其进行分区加权求二次帧差,从而确定镜头变换所在的GoP;在GoP中通过计算P帧和B帧的不同类型宏块比率,来精确定位镜头边界。实验结果表明,该方法可大大减少计算时间和数据量,且具有良好的检测效果。Aimed at the shortcoming in conventional airspace shot detection need to decompress, computationally intensive and low efficiency, a video shot detection method based on compressed domain is proposed. In this method, the 8 low-frequency DCT coefficients of I frames were extracted first from the video stream according to MPEG standard, and the frames are partitioned, weighted and computed the twice-frame-difference, so the CoP of the shot boundary is determined ; In the CoP, by calculating the different types of macro-block ratios of the P and B frames, to accurately detect shot boundary. Experiment result showed that, the amount of computation and data is reduced greatly and has a good detection results.

关 键 词：视频镜头检测 压缩域 MPEG 二次帧差 宏块比率 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]